A data-driven mechanistic approach to IPF target discovery
As IPF progresses, patients often need oxygen and, in some cases, lung transplantation. The prognosis can be devastating, and there is a clear unmet need for better treatments. One of the challenges in IPF target discovery is selecting which biological mechanism to target; we want to treat the mechanisms causing the disease in order to inhibit disease progression. As the name suggests, the cause of IPF is idiopathic, or unknown, and the exact mechanisms involved in the pathogenesis of IPF are still unclear.
Our collaboration with AstraZeneca aims to address this challenge by leveraging AI and machine learning to prioritise and select the right mechanism to target in the right patient population.
Knowledge: building the strongest possible data foundation
Traditionally, IPF scientists search for clues manually in a growing corpus of data that is simply too vast to keep up with or comprehend. This data is usually analysed in silos, which obscures the full picture of disease networks and often leads to failure down the line of development. We solve this problem by aggregating disparate data modalities into our Biomedical Knowledge Graph using AI and machine learning. In the IPF project, we enriched the Knowledge Graph with IPF specific data including omics and AstraZeneca’s proprietary data, to give scientists a window into all available biomedical data, without boundaries or bias.
Target identification: leveraging AI to make novel target predictions
We then use our powerful machine learning models to query that data at scale, empowering scientists to explore novel connections and previously unknown areas of biology to identify multiple potential novel targets for IPF.
Target triage: augmenting scientific expertise
Our AI-assisted target triage tools help our drug discoverers contextualise the suggested targets by presenting the scientific evidence underpinning the predictions. In doing so, scientists can make informed data-driven decisions over which targets to prioritise.
Precision medicine: increasing the likelihood of success
IPF is a heterogeneous disease; progression and clinical endpoints vary between patients. We use machine learning algorithms on patient-level data to identify specific groups that share common features and then focus on novel drug targets that may be key drivers of disease in each group. This precision medicine approach increases the likelihood that a drug will succeed in clinical trials.